Meta’s US$2 billion acquisition of AI agent startup Manus is a lesson in cross-border compliance risks for corporate CTOs everywhere. China’s Ministry of Commerce announced on January 9 that it would investigate whether the deal violates export controls, technology transfer rules, and foreign investment regulations, even though Manus will move from Beijing to Singapore in 2025.
This research reveals an uncomfortable reality for enterprise AI buyers. A vendor’s headquarters location tells you nothing about regulatory risk.
“The AI agent that Manus developed was definitely one that could be subject to export controls by Chinese regulators,” said Dai Menghao, a Shanghai-based partner at King & Wood Mallesons, which specializes in export controls and sanctions. South China Morning Post. It is technology, not business registration, that determines jurisdiction.
When transfers do not equate to regulatory freedom
Manas appears to have ticked all the boxes for regulatory independence. The company will relocate its 105-person team from Beijing to Singapore in summer 2025, lay off 80 mainland employees, set up offices in Singapore, Tokyo and San Francisco, and secured US$75 million in US funding from Benchmark.
“Following the transaction, there will be no continued Chinese ownership of Manus AI, and Manus AI will discontinue its services and operations in China,” Meta said in a statement in December.
However, Ministry of Commerce spokesperson He Yadong clarified that corporate structure alone does not determine compliance. “The Chinese government consistently supports enterprises in carrying out mutually beneficial cross-border business and international technical cooperation in accordance with laws and regulations,” he said at a press conference on January 9.
“However, it should be noted that enterprises’ external investments, technology exports, data exports, and cross-border acquisitions must comply with Chinese laws and regulations and undergo due process.”
Cui Fan, a professor at the International University of Business and Economics and chief expert at the China World Trade Organization Research Association, said the investigation will look into when, how and which technologies Manus transferred overseas from companies based in China.
The company’s founders could face criminal charges under Chinese law if regulators decide that Manus should have obtained an export license before transferring technology or talent.
The regulatory framework that corporate buyers must understand
China updated its technology export control regulations in 2020, expanding the scope to include certain algorithms. The change is widely interpreted as giving the Chinese government a stronger legal basis to intervene in transactions involving strategic technology.
The update gained attention after the U.S. pressured ByteDance to sell TikTok’s U.S. operations and China asserted its authority over foreign technology transfers. This framework covers three key areas that enterprise AI buyers should understand when assessing vendor risk.
Export control: Advanced AI agents, models, and related intellectual property qualify as strategic assets subject to licensing requirements. Regardless of where the company later incorporates, the Chinese government will maintain jurisdiction over technology developed in China.
Data security rules: Cross-border data transfers require regulatory approval, especially for datasets used to train and fine-tune AI models. Where the training takes place is more important than where the inference takes place.
Foreign investment regulations: When Chinese people transfer technology assets overseas, even through legal corporate restructuring, authorities will assess whether the transfer requires government permission.
Wang Yimin, a partner at Beijing Xinzheng Law Firm, estimates that the Manus review could take up to six months, consistent with the timeline for similar technology transfer evaluations. “This could be a high-profile test case for China’s equivalent of the Committee on Foreign Investment in the United States,” said Winston Ma, an adjunct professor at New York University School of Law who specializes in AI and the digital economy. SCMP.
What this means for AI vendor due diligence
The Manus case exposed gaps in the way enterprise buyers assess the regulatory risk of AI vendors. Standard procurement processes focus on data residency, service level agreements, and contractual responsibilities.
Few assess whether a vendor’s history of technology development puts them at risk of ongoing compliance in multiple jurisdictions.
Enterprise buyers should ask AI service providers:
Questions about the origin of technology:
- Where was the core AI model or agent developed?
- Which jurisdiction’s export control regime can claim authority?
- Are any of your team members involved in training Chinese people?
Transfer compliance:
- If the company relocated, what regulatory approvals were obtained?
- Can the vendor demonstrate compliance with export licenses for technology transfer?
- What contingencies arise if regulators challenge past transfers?
Continuity of operations:
- How will regulatory investigations impact service delivery?
- What customer notification obligations do I have during the review period?
- Does the vendor maintain insurance and reserves for regulatory risks?
“The most likely outcome I think is not an outright block, but a longer approval process and potential conditions on how Manus technology developed in China can be used,” Nick Patience, head of AI at Futurum Group, told CNBC. “However, the threat of stricter measures gives Beijing bargaining power in high-profile US-led acquisitions.”
Upfront risks for enterprise AI strategies
This investigation is important beyond Meta’s specific contract. If the Chinese government determines that it can effectively assert jurisdiction over Chinese-originated AI technologies regardless of corporate restructuring, it would establish a precedent for continued regulatory enforcement of companies’ AI supply chains.
Corporate buyers using AI agents for market research, coding assistance, or data analysis (exactly what Manas was offering before the acquisition of Meta) are now facing questions about the provider’s stability during geopolitical conflicts. The company reached $100 million in annual recurring revenue within eight months of launch, demonstrating rapid enterprise adoption and how mission-critical dependencies can form quickly.
Winston Ma said the smooth approval could “create new avenues for young Chinese AI startups,” which could include a combination of physical relocation and overseas acquisitions to circumvent restrictions on technology transfer.
Conversely, the regulatory intervention suggests that the Chinese government will continue to pursue Chinese-originated AI companies even after they are relocated, potentially closing off what seemed like an escape route for startups to overcome U.S.-China tensions.
The lesson for enterprise AI buyers is to recognize that compliance risks for AI vendors extend beyond contract terms to nebulous jurisdictional questions about where and by whom the technology was first developed. This is a due diligence requirement, but most procurement teams have not yet built the capacity to assess it.
See also: Manus AI Agent: A breakthrough in China’s agent AI
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California, and London. This comprehensive event is part of TechEx and co-located with other major technology events. Click here for more information.
AI News is brought to you by TechForge Media. Learn about other upcoming enterprise technology events and webinars.
